|
Abstract
Last Friday, we discussed the memory boundness problem of evaluating
continuous queries in a data stream system. It was assumed that the
answers to the queries are exact i.e., no errors or approximations to
the answers are allowed. In this talk, we relax this requirement.
We investigate how, by sacrificing certain degree of accuracy,
continuous queries can be executed efficiently in terms of
space and time.
Specifically, we introduce the technique of data sketching. We will
see how data sketching is applied to answer aggregate queries (e.g.,
self-join size, multi-join size) efficiently with a high probability
of getting a small error.
Read the Presentation
Slides...
Referred Papers
|